Sample Design

1
Sample Design

• Michael DeBacker
• and Lloyd Morrison

2
Important Underlying Themes of Sample Design

• Ability to make inferences
• Whether or not to stratify
• Co-location/co-visitation

3
Inference

• We can only sample a very small proportion (often
  lt1) of most natural areas, but
• Our job is to protect, restore, understand, and
  inform others about the entire area, not just
  some convenient portion of it.
• We need to make scientifically defensible
  inferences to areas beyond the limited sites we
  sample.
• Statistical, design-based inferences can only be
  made to areas that have a chance of being
  included in the sample.

4
Stratification

• Stratification is a powerful method of decreasing
  variance among sample sites, but
• Strata must be carefully chosen in long-term
  studies, so criteria for strata do not change.
• Multiple protocols may need to use different
  criteria for stratification.
• Later addition of protocols may not fit within
  defined strata.
• In general, stratification is not recommended for
  long-term studies measuring multiple response
  variables.

5
Co-location (Co-visitation)

• Co-location will provide information on multiple
  vital signs for the same sites.
• Co-visitation will aid overall sampling
  efficiency, as multiple vital signs can be
  sampled at the same time.

6
Status of an Overall Sampling Framework for the
First Twelve Networks
7
Summary of Spatial Designs Utilized by the First
Twelve Networks
8
Example 1 Integrated Terrestrial
DesignHeartland Network Prairie Cluster
Prototype

• Protocols Spatially Integrated
• Vegetation communities
• Bird communities
• Invasive plants
• Protocols Temporally Integrated
• None
• Strata
• soils, slope aspect (vegetation monitoring only)

9
Monitoring projects are integrated using an
underlying grid for multiple purposes Step
one overlay the sample frame with a relatively
fine scale grid, the vertices of which create a
systematic sample.
10
Invasive plant monitoring - rapid data
collection - maximum spatial coverage - all
points in the reference frame are sampled
Invasive plant species (INP) sample site
11
Protocols for monitoring breeding birds utilize a
systematic survey design at a larger spatial
scale. In this example, the initial grid would
be sub-sampled to the desired scale as indicated
by blue circles.
Invasive plant species (INP) sample site INP and
bird community sample site
12
For vegetation community monitoring, the grid
vertices form a pool of potential sample sites
from which a stratified random sample is drawn.
Invasive plant species (INP) sample site INP and
bird community sample site INP and vegetation
community sample site
13
Example 2 Systematic, Unequal Probability
Sampling Design Northern Colorado Plateau Network
(NCPN)

• Protocols Spatially Integrated
• Vegetation
• Soil/Site Stability
• Hydrologic Function
• Soil Crust Structure
• Nutrient Cycling
• Protocols Temporally Integrated
• TBD
• Probability of site inclusion determined by
• Accessibility
• Ecological Sites

14
Accessibility Model Example from Zion National
Park Selection probabilities for sample plots
are defined by accessibility (high or low)
15
Selection probabilities for sample plots are also
determined by ecological sites Ecological
sites are defined by climate, geology, and soils.
16
Sample grid overlaid on an Ecological Site map
A final sample is drawn from this pool of points
utilizing the inclusion probabilities.
17
Example 3 Two Stage Systematic Sampling,
Central Alaska Network (CAKN)

• Protocols Spatially Integrated
• Vegetation
• Passerine Birds
• Snow depth
• Protocols Temporally Integrated
• Vegetation and Passerine Birds (50)
• Strata
• Road corridor at DENA

18
Stage one - a systematic grid, 20 x 20 km Stage
two a 25 point mini grid, 500 x 500 m
19
Benefits of the two-stage systematic grid design

• concentrates landscape-scale sampling effort
  within study areas

- access cost per data point is lower - fewer
trips into wilderness are required per sample

• effectively samples both regional and meso-scale
  gradients in resource conditions

• relies on a sampling frame that is not tied to
  any preconceived
• notions of how changes in the ecosystem
  will occur

• provides a multiple-scale sampling frame that
  allows for
• nesting of monitoring efforts that occur
  at different spatial scales or at different
  levels of intensity

• if funding is interrupted prior to completion of
  the full park-wide sample, partial data set
  remains a valid network of monitoring sites with
  interpretable data and a subset of sites may be
  continued into the future

20
Temporal Design for Vegetation and Snow Depth
Monitoring Projects
21
Temporal Design for the Passerine Birds
Monitoring Project
Group A sites near the main road
Group B sites dispersed throughout parks,
provides statistical connectivity
Group C sites dispersed throughout parks,
provides broad spatial coverage
22
Example 4 GRTS Design for Aquatic Resources
Monitoring Heartland Network (HTLN)

• Protocols Spatially Integrated
• Fish Communities
• Water Chemistry
• Aquatic Invertebrates
• Geomorphology
• Ozark Hellbender (TE)
• Protocols Temporally Integrated
• Fish, Invertebrates, and Water Chemistry
• Strata
• None

23
Ozark National Scenic Riverways Stream Network
508 KM
GRTS spatially-balanced sample, n 50
24
(No Transcript)
25
Temporal Design for Hellbender, Fish,
Geomorphology, and Aquatic Invertebrate
Monitoring Projects
26
Acknowledgments

• Trent McDonald (West, Inc.),
• Tony Olsen (EPA),
• Paul Geissler (USGS),
• and others.